Abstract

We present a methodology for the algorithmic construction of Lyapunov functions for the transient
stability analysis of classical power system models. The proposes methodology uses recent advances in
the theory of positive polynomials, semidefinite programming, and sum of squares decomposition,
which have been powerful tools for the analysis of systems with polynomial vector fields. In order to
apply these techniques to power grid systems described by trigonometric nonlinearities we use an
algebraic reformulation technique to recast the systems dynamics into a set of polynomial differential
algebraic equations. We demonstrate the application of these technique to the transient stability
analysis of power systems by estimating the region of attraction of the stable operating point. An
algorithm to compute the local stability Lyapunov function is described together with an optimization
algorithm designed to improve this estimate.
I. INTRODUCTION
A traditional approach to transient stability analysis of power systems involves the numerical
integration of the nonlinear differential equations describing the complicated interactions between its
components. This method provides an accurate description of

transient phenomena but its

computational cost prevents time domain simulations from providing real time transient stability
assessments and significantly constraints the number of cases which can be analyzed [1].
Alternative approaches to transient stability analysis have bee intensively explored [1] [5].
Among the methods proposed, the so called direct methods avoid the expensive time domain
integration of the postfault system dynamics. These methods rely on the estimation of the stability
domain of the postfault equilibrium point. If the initial state of the postfault system lies inside the
stability domain, then we can assert without numerically integrating the postfault trajectory that the
system will eventually converge to its postfault equilibrium point. This inference is made by
comparing the value of a carefully chosen scalar state function (energy and Lyapunov functions) at the
clearing time to a critical value. In practice, finding analytical energy and Lyapunov functions for
transient stability analysis has encountered significant difficulties. For example, the energy function
approach to transient stability analysis relies on two strong requirements. First, we should be able to
define an analytic energy function. This condition is generally violated in practice since energy
functions for power systems with transfer conductances do not exist [5], [6]. Thus, for systems with
losses, no analytical expressions are available for the estimated stability boundary of the operating
point. Second, we should stability compute the critical energy value. This task is also very difficult and

can provide inaccurate stability assessments if it returns the wrong critical value [7]. The closest
Unstable Equilibrium Point (UEP) method provides sufficient but not necessary conditions for stability
and is conservative. This method requires the identification of all equilibrium points located on the
boundary of the stability region. This requires a significant computational effort and it is impractical,
but it offers mathematical guarantees. The controlling UEP provides less conservative estimates of the
stability boundary than the closest UEP. It is generally very difficult to find the controlling UEP
relative to the fault-on trajectory [7]. Nevertheless, a systematic method called the boundary of
stability region based controlling UEP method (BCU method) has been developed to find this point
[8], [9]. Extensive numerical simulations have found counter-examples [10] where the BCU method
fails to give the correct answer, predicting stability for systems that in fact suffer from second-swing
instability. Furthermore, it has been shown that the mathematical assumptions of the BCU method do
not hold generically and that the theoretical guarantees for the BCU method are, at best, questionable
[6], [11].
On the other hand the Lyapunov function approach to transient stability analysis has been
traditionally considered very difficult due to the lack of a systematic methodology for constructing a
Lyapunov function see [12]-[14] for details and a systematic survey of Lyapunov functions in power
system stability. The method of Zubov is an exception and, in principle, can find a Lyapunov function
and determine the exact boundary of the Region Of Attraction (ROA). This method requires the
solution of a Partial Differential Equation (PDE) which does not possess in general a closed form
solution. Moreover, for power system models, the existence of transfer conductances has proven
again to be a serious difficulty. This is the case, for example, when using the multivariable Popov
stability criterion. This method can also construct a genuine Lyapunov function, but requires the
satisfaction of sector conditions that break down in the presence of transfer conductances see, for
example, [15]-[17] and references therein.
More recent results in the literature, using a passivity-based control methodology (see [18] and
references therein), show the existence of Lyapunov functions for small, but unspecified, transfer
conductance and require the solution of a formidable system of PDEs (additionally, the angle
differences in equilibrium are also required to be small). Another recent result [19], which is close
methodologically to our approach, estimates the ROA for non-polynomial systems using truncated
Taylor expansions and semidefinite programming optimizations see also [20] for a comprehensive
description of Sum OF Squares (SOS) programming techniques for the estimation and control of the
domain of attraction. They cannot be used in transient stability assessment or in estimating the critical
clearing time. The method in [22] proposes a procedure to construct Lyapunov functions for power

with transfer conductance using dissipative systems theory for large scale interconnected systems. This
approach is the only one that we found in the literature where the condition of small transfer
conductances is not necessary. Nevertheless, it still contains some restrictive sector conditions on the
nonlinearities which translate into conditions on the angle differences in equilibrium. It also contains
many parameters that have to be finely tuned in order for the method to converge. The method in [23]
uses an extension of LaSalles Invariance Principle to find extended Lyapunov functions for power
systems with transfer conductances. The derivative of the extended Lyapunov function is not required
to be always negative semidefinite and can take positive values in some bounded regions. This is a
very interesting and promosing approach. Moreover, the authors propose a generic Lyapunov function
for multimachine systems. The conditions in the Extended Invariance Principle require that the transfer
conductances be small in order for the domain in which the derivative is positive to be include in the
bounded domain defined by the Lyapunov function. Usually, these domain inclusions are very difficult
to compute numerically and the assumption that the transfer conductances are small is necessary in
order to guarantee these conditions.
The main contribution of this paper is twofold. First, we introduce an algorithm that constructs
Lyapunov functions for classical power system models. Seconds, we embed this algorithm into an
optimization loop which seeks to maximize the estimate of the region of attraction of the stable
operating point. Our approach exploits recent system analysis methods that have opened the path
toward the algorithmic analysis of nonlinear systems using Lyapunov methods [24]-[30]. We introduce
three critical steps that are necessary in this formulation. For dynamical systems described by
polynominal vector fields, the first step is to relax to the non-negativity conditions in Lyapunovs
theorem to appropriate Sum Of Squares (SOS) conditions which can be tested efficiently using
semidefinite programming (SDP) [24]. The SOS technique cannot be applied directly to power grid
systems since they are not defined by polynomial vector fields. Hence, the second step is to generalize
the SOS formulation to non-polynomial systems using a procedure which recasts the original nonpolynomial system into a set of polynomial differential algebraic equations [27]. Finally, since the
recasted system evolves over algebraic equality constraints, we employ a fundamental theorem from
real algebraic geometry [31] in order to provide a convex relaxation of the quality and inequality
conditions required by Lyapunovs theorem in this case [25]. The proposed algorithm is used to find
Lyapunov functions and estimates of the Region Of Attraction (ROA) for two power system models.
We formulate an optimization algorithm that searches over the space of polynomial Lyapunov function
in order to improve these estimates. For the power system model without transfer conductance we
compare the performance of the proposed algorithm to the energy function method. We apply the same
analysis to the power system with transfer conductances for which an exact energy function does not

exist but for which a Lyapunov function has been proposed in the literature. A critical discussion of the
method is also presented. Extensions and a discussion of the steps required to generalize this analysis
to large scale systems are also described. The SOS programming concepts introduced in this paper are
not new but, to the best of our knowledge, they have never been applied to the transient stability
analysis of power systems.